A Transfer Learning Algorithm Based on Regularized Discriminant Analysis

Aiming at the problem that most instance-based transfer learning methods are difficult to estimate the distribution parameters and having poor generalization ability, a regularized discriminant transfer learning algorithm is proposed. Based on the discriminant analysis and semi-supervised learning t...

Full description

Saved in:
Bibliographic Details
Main Authors: WANG Li-li, FENG Qi-shuai, CHEN De-yun, YANG Hai-lu
Format: Article
Language:zho
Published: Harbin University of Science and Technology Publications 2019-04-01
Series:Journal of Harbin University of Science and Technology
Subjects:
Online Access:https://hlgxb.hrbust.edu.cn/#/digest?ArticleID=1664
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:Aiming at the problem that most instance-based transfer learning methods are difficult to estimate the distribution parameters and having poor generalization ability, a regularized discriminant transfer learning algorithm is proposed. Based on the discriminant analysis and semi-supervised learning theory, the semi-supervised Gauss kernel discriminant analysis method is studied by kernel method and regularization method, and the reusable samples are transferred by constructing the revised embedding space. On the one hand, screening samples in the mapping space can solve the difficulty of estimating the parameters of domain distribution; on the other hand, introducing pseudo-labeled data and defining the distance function can avoid over-fitting problems. The experimental results on text and non-text datasets validate that the proposed algorithm can effectively improve the accuracy and generalization ability of transferring.
ISSN:1007-2683